Nuclear power has been a stagnant U.S. energy source for decades, but it could be primed for a technology-driven renaissance amid the growing artificial intelligence boom.
I get your point, but have you looked into the power demands of data centers? They already have room filling batteries for power outages, but those are just enough to keep the lights on while the diesel generators start.
There are immense capacity utility scale batteries available now from dozens of vendors. They would be roughly 100x easier to build than a nuclear power plant, even with a solar farm attached.
The most recent nuclear power plant built in the US was the 2 new units at the Vogtle plant in Georgia. They took 11 years and 34 billion dollars to build to output a roughly 2.4 gw of steady power.
A 1gw solar + battery plant was built in Nevada that cost 1.9 billion. They secured financing in 2022, and finished building it in 2024.
So we can get a solar array built to do the above with battery storage for 4 billion, in 2-4 years. For the same cost as that added 2.4GW nuclear, we could build 18GW of solar with 12.6gwh of storage.
So nuclear will do 2.4GW of peak, with 2.4gwh of “storage” available 24/7.
I have no doubt that the above 18GWh of solar could be traded in for more battery, to a more sane ratio that could compete with that “storage” while also providing 4-5x of total power. Based on what I can find, the batteries were about 1/2 the cost, so if we knock the solar generation to 9GW, we can increase battery to 25.2GWh. Now you still have huge power generation, a huge power storage that you can use all at once or over a long period, and it matches the “storage” that that nuclear plant offers for 12hrs, I.e the time when the sun is down.
It’s honestly baffling why these companies are trying to spin up nuclear plants instead of pushing ahead with more grid renewables.
This is just the peak power. The average power is much less. And batteries can maybe work on a grid scale for smoothing, but not for an individual consumer like a data center.
I get your point, but have you looked into the power demands of data centers? They already have room filling batteries for power outages, but those are just enough to keep the lights on while the diesel generators start.
There are immense capacity utility scale batteries available now from dozens of vendors. They would be roughly 100x easier to build than a nuclear power plant, even with a solar farm attached.
The most recent nuclear power plant built in the US was the 2 new units at the Vogtle plant in Georgia. They took 11 years and 34 billion dollars to build to output a roughly 2.4 gw of steady power.
A 1gw solar + battery plant was built in Nevada that cost 1.9 billion. They secured financing in 2022, and finished building it in 2024.
So we can get a solar array built to do the above with battery storage for 4 billion, in 2-4 years. For the same cost as that added 2.4GW nuclear, we could build 18GW of solar with 12.6gwh of storage.
So nuclear will do 2.4GW of peak, with 2.4gwh of “storage” available 24/7.
I have no doubt that the above 18GWh of solar could be traded in for more battery, to a more sane ratio that could compete with that “storage” while also providing 4-5x of total power. Based on what I can find, the batteries were about 1/2 the cost, so if we knock the solar generation to 9GW, we can increase battery to 25.2GWh. Now you still have huge power generation, a huge power storage that you can use all at once or over a long period, and it matches the “storage” that that nuclear plant offers for 12hrs, I.e the time when the sun is down.
It’s honestly baffling why these companies are trying to spin up nuclear plants instead of pushing ahead with more grid renewables.
This is just the peak power. The average power is much less. And batteries can maybe work on a grid scale for smoothing, but not for an individual consumer like a data center.
These are grid scale projects. Each one is designed to power 100s of thousands of homes.
Based on the data I can find, 35MW datacenters are what AI will require by 2030. The project I linked above could power 20+ of them at once by itself.